Correlation between breeding values for milk fatty acids and Nordic Total Merit index traits for Danish Holstein and Danish Jersey

Milk fatty acid composition is gaining interest in the Danish dairy industry both to develop new dairy products and as a management tool. To be able to implement milk fatty acid (FA) composition in the breeding program, it is important to know the correlations with the traits in the breeding goal. To estimate these correlations, we measured milk fat composition in Danish Holstein (DH) and Danish Jersey (DJ) cattle breeds using mid-infrared spectroscopy. Breeding values were estimated for specific FA and for groups of FA. Correlations with the estimated breeding values (EBV) underlying the Nordic Total Merit index (NTM) were calculated within breed. For both DH and DJ, we showed that FA EBV had moderate correlations with the NTM and production traits. For both DH and DJ, the correlation of FA EBV and NTM were in the same direction, except for C16:0 (0 in DH, 0.23 in DJ). A few correlations differed between DH and DJ. The correlation between claw health index and C18:0 was negative in DH (−0.09) but positive in DJ (0.12). In addition, some correlations were not significant in DH but were significant in DJ. The correlations between udder health index and long-chain FA, trans FA, C16:0, and C18:0 were not significant in DH (−0.05 to 0.02), but were significant in DJ (−0.17, −0.15, 0.14, and −0.16, respectively). For both DH and DJ, the correlations between FA EBV and nonproduction traits were low. This implies that it is possible to breed for a different fat composition in the milk without affecting the nonproduction traits in the breeding goal.


INTRODUCTION
In the Danish dairy industry, there is increasing interest in developing new dairy products with fat as an important component. Although the cows' diet, along with other nongenetic factors like parity and lactation stage, has a large influence on the variation (Hein et al., 2016), there is substantial genetic variation for individual fatty acid (FA) content in milk (Krag et al., 2013).
The gold standard for measuring milk fat composition is based on GC; however, this method is time consuming and costly. Thus, GC is not the method of choice when large quantities of samples must be analyzed. To be able to provide the Danish dairy industry with routine milk fat composition data, mid-infrared spectroscopy (MIRS) prediction models have been applied (Fleming et al., 2018;Hein et al., 2018). Hein et al. (2018) showed that mid-infrared spectroscopy was useful for collecting large quantities of detailed milk FA profiles from individual cows. The data showed that there was genetic variation for a variety of fat components and specific FA, suggesting that changing the milk FA profile using genetic selection would be possible (Hein et al., 2018). These data provided the base for EBV for FA composition, as shown by Poulsen et al. (2020). However, before milk fat composition can be implemented in the breeding goal, genetic correlations with the other traits in the breeding goal must be known to avoid selecting for FA with unfavorable correlations with other breeding goal traits.
Since 2015, milk samples that are part of the Danish milk recording are screened for their specific milk FA using the commercial software of the Foss Application Note 64 (FOSS). Therefore, the aim of this study was (1) to evaluate the correlations between the milk FA profile, as provided by the Foss Application Note 64, and the breeding goal traits by EBV for 4 individual FA and 7 FA groups in Danish Holstein (DH) and Danish Jersey (DJ), and (2) to estimate correlations between EBV for the individual FA and FA groups with EBV for index traits in the Nordic Total Merit index (NTM). Specific focus was placed on correlations

Correlation between breeding values for milk fatty acids and Nordic Total Merit index traits for Danish Holstein and Danish Jersey
between the fat index, which is a combination of SFA, PUFA, and MUFA, and the 4 individual FA and 7 FA groups for investigating the loss in overall fat gain when selecting for a specific FA profile in the milk.

MATERIALS AND METHODS
No animal experiments were performed in this study, and therefore, approval from the ethics committee was not required. The total amount of FA is the sum of the SFA, MUFA, and PUFA fractions or the sum of the SCFA, MCFA, and LCFA fractions. In Figure 1, the relationships between the different FA fractions are shown: SCFA, MCFA, and LCFA consist of some SFA; MUFA is part of both MCFA and LCFA; and PUFA is part of LCFA. The individual FA C14:0 and C16:0 are included in both the MCFA and SFA fractions; C18:0 is part of the SFA fraction as well as of the LCFA fraction; C18:1 is part of LCFA and MUFA fractions. The TFA group of FA is part of the LCFA, MUFA, and PUFA fractions ( Figure 1).

Data Editing of FA Phenotypes.
Data editing of the FA phenotypes is described in detail by Hein et al. (2018). In short, only first-parity cows were selected. Thereafter, the following editing procedures were performed: (1) if one or more FA traits were missing from an observation, the observation was removed; (2) if the PUFA concentration was greater than or equal to the MUFA concentration, the observation was removed; (3) if the ratio of the sum of SFA, MUFA, and PUFA to the fat percentage was <0.825 or >1.075, the observation was removed; (4) if the ratio of the sum of SCFA, MCFA, and LCFA content to the sum of the SFA, MUFA, and PUFA content was <0.84 or exceeded >1.04, the observation was removed; (5) if the fat percentage was >8% for DH or >12% for DJ, the observation was removed. The lactation period was defined as a period between 8 and 305 DIM. In addition, daughters of sires with <10 daughters with FA phenotypes and cows that changed herd during the sampling period were removed from the data set (Hein et al., 2018). After editing, data consisted of 425,559 first-parity DH cows with a total of 2,135,955 FA test-day observations and 70,461 firstparity DJ cows with 333,455 FA test-day observations. Pedigrees were traced back for 3 generations, resulting in a total number of animals of 1,277,430 in the DH pedigree, and of 124,508 in the DJ pedigree.

Breeding Value Estimation
Fatty Acids. Breeding values were estimated using the DMU software package (Madsen and Jensen, 2007). The breeding values were estimated within breed for each FA fraction and total fat content using the following univariate repeatability linear animal model: where y is the vector of test-day observations, b is the vector of fixed effects, h is the vector of random herd effects, a is the vector of random animal effects, pe is the vector of permanent environmental effects, e is the vector of residual effects, and X, Q, Z, and W are incidence matrices relating records to fixed, herd, animal, and permanent environmental effects, respectively. Fixed effects were month × year of recording and a fixed regression accounting for the variability in early lactation determined by fitting a lactation curve with the following function (Wilmink, 1987): where y t is the FA content in grams of FA/sum of the SFA, MUFA, and PUFA content or in grams of FA/100 grams of milk at t DIM, and a, b, and c are regression coefficients. The parameters jointly describe the shape of the lactation curve: a sets the amplitude of daily FA content, b is the production increase toward the peak in early lactation, and c is the linear decline after the peak, t is the time in DIM, and k affects the duration of the acceleration period (fixed at k = 0.05; Wilmink, 1987).
The breeding values for FA were standardized with mean of 100 and a standard deviation (SD) of 10 for the 1,469 Holstein bulls and 303 Jersey bulls with NTM and daughters with phenotypic FA data according to model [3]: where EBV is the estimated breeding value, EBV is the mean EBV, and meansd is the standard deviation of the EBV for the animals in the base population, which consists of sires with NTM and at least 10 daughters with FA observation. For the individual FA and FA groups, a higher EBV always indicates higher values of the FA. A higher EBV can be unfavorable if a FA is unfavorable for human health, and the content of the FA is therefore preferred to be lowered. Accuracies of the breeding values for the different milk FA in DH and DJ are presented in Table 1.

EBV for Index Traits in NTM.
The following indices were considered: yield, growth, fertility, birth (direct), calving survival (maternal), udder health, general health, frame, feet and legs conformation, udder conformation, milkability, temperament, longevity, claw health, youngstock survival, and milk, fat, and protein production. The phenotypes underlying the indices were described in detail by NAV (2020) and are presented in Table 2. Estimated breeding values from the Nordic genetic evaluation (NAV) in November 2018 were used. The breeding values for the index traits were estimated as described in NAV (2020); for these indices, a higher value is always favorable. In total, there were 2,039 Holstein sires (of which 1,469 had NTM values) and 2,014 Jersey sires (of which 303 had NTM values).

Statistical Analysis
Correlations between the breeding values for the sires were estimated using the PROC CORR procedure in SAS (version 9.4; SAS Institute Inc.).

Milk FA Phenotypes
The mean values of the milk FA for the 2 breeds are presented in Table 3. On average, the tendency was that DH had lower SFA and higher MUFA and PUFA than DJ. In contrast, DH had lower SCFA and MCFA but higher LCFA than DJ. For the 4 individual milk FA, DH showed lower content of C14:0 and C16:0, but higher C18:0 and C18:1 in milk compared with the DJ (Table 3).

Correlations Between FA EBV and NTM Breeding Goal Traits
The EBV for FA were either based on percentage of the sum of SFA, MUFA, and PUFA or grams/100 grams of milk. The correlations between the FA EBV based on Buitenhuis et  percentage of the sum of SFA, MUFA, and PUFA and the different indices of the NTM breeding goal traits are presented in Table 4 for DH and in Table 5 for DJ.
In general, correlations for DH between the FA indices and index trait indices were low (<|0.2|), except for the milk production-related indices (yield, milk, fat) and NTM. The EBV for SFA, SCFA, and MCFA showed positive correlations with yield index (0.29, 0.29, and 0.24, respectively) and fat index (0.43, 0.38, and 0.30, respectively) but negative correlations with milk index (−0.20, −0.10, and −0.26, respectively) and positive correlations with NTM (0.29, 0.29, and 0.24, respectively). Correlations with yield, milk, fat, and NTM were generally lower with PUFA than with SCFA and were higher with MCFA than with LCFA (Table 4).
Correlations between EBV for protein index and milk FA EBV were either low or not different from zero, except for C16:0 (−0.13) and C18:0 (−0.08). The EBV for MUFA showed significant but numerically low negative correlations with the 3 health-related indices udder health (−0.12), general health (−0.11), and claw health (−0.11), whereas EBV for SFA, SCFA, and C14:0 with the 3 health indices also showed significant, numerically low but positive correlations (Table 4).    In general for DJ, many of the correlations were not significant from zero, and significant correlations (different from zero) were moderate (<|0.4|). The significant correlations were mainly between FA indices and production trait indices (yield, milk, fat) and NTM, but udder health and general health indices also showed moderate correlations with some of the FA indices (Table 5).
The correlations between the FA EBV based on grams/100 grams of milk and the different index EBV are presented in Supplemental In general, for both DH and DJ, we found positive correlations between FA indices and yield, growth, frame, milkability, temperament, milk, protein, fat, and NTM indices. Negative correlations between FA indices and fertility, calving survival, udder health, general health, feet and legs, udder, longevity, and claw health indices were found for DH (Supplemental Table S1). For DJ, positive correlations between FA indices and yield, birth calving survival, milkability, temperament, longevity, milk, protein, fat, and NTM indices were found. Negative correlations between fertility, udder health, general health, feet and legs, and udder indices and the FA indices were found (Supplemental Table S2).

DISCUSSION
To change milk fat composition via breeding, information about genetic correlations between the new FA traits and the breeding goal traits is vital. Recent stud-ies have estimated correlations among mid-infraredpredicted FA (Hein et al., 2017;Lopez-Villalobos et al., 2020), and correlations between mid-infrared-predicted FA and milk production traits (Fleming et al., 2018) used direct genetic correlations between the traits. Not much is known about the relation of mid-infrared-predicted FA and breeding goal traits. The approximate genetic correlations, as estimated in our study, between breeding values for milk FA traits and those of index traits underlying the NTM index may provide a first indication of the mutual relationship between milk FA profile and the breeding goal.
Selection for total fat in dairy cattle increases the amount of SFA in the fat composition, as the genetic correlation between SFA and total fat is positive (Hein et al., 2017). Indeed, Eskildsen et al. (2017) stated that the measurement of individual milk FA by mid-infrared is indirect and primarily depends on the covariation between the predicted FA and total fat. However, in our previous study, mid-infrared-predicted SFA showed a clear positive genetic correlation with total fat, whereas MUFA and PUFA showed negative genetic correlations with total fat and SFA (Hein et al., 2017), indicating that the distribution of SFA, MUFA, and PUFA within total fat can be affected by the correct weighting of the traits.

Changing the Milk FA Profile
Whether milk composition should be adapted to better meet human nutritional requirements has been a topic of discussion for some time (Givens, 2010). In the case of milk fat, this generally implies milk with a higher content of unsaturated fat. It may also be beneficial to change the ratio between the different FA groups or specific FA in the total fat content. This can be achieved either by changing the feeding regimen (e.g., Poulsen et al., 2020) or using genetics, as it has been shown that there is a genetic component underlying milk FA groups and individual milk FA (Krag et al., 2013;Hein et al., 2018). Using genetic selection to change the specific milk FA composition has been difficult because the current gold standard for measuring specific FA by GC is labor intensive and expensive, and these studies have therefore been restricted in the number of animals and samples available (Soyeurt et al., 2006). When infrared methods were developed for milk FA (Soyeurt et al., 2006), this opened the possibility to measure the detailed milk profile on a larger scale. Since 2017, the detailed milk FA profile of the Danish cattle population has been measured using the Foss Application Note 64 method (Hein et al., 2018).
The amount of collected data is beneficial for estimation of accurate breeding values for infrared-predicted FA breeding values. However, the accuracy of the current approach is limited by the accuracy of the infrared prediction, as provided by the external cross validation R 2 of the infrared prediction equations. In this study, we showed low to moderate correlations between different milk FA and, especially, traits related to yield. Interestingly, correlations different from zero were also detected between milk FA and health traits, such as udder health and general health. This was more pronounced in DJ than in DH. This indicates that, if desired, there is room to improve the FA profile in milk without hampering current breeding goal traits. However, it is necessary to calculate economic values for appropriate weighting of the individual FA and FA fractions to favor the desired FA profile and avoid unwanted side effects such as lower yield or increased disease frequencies. An alternative "Healthy Fatty Acid" (HFA) index is therefore suggested. Using the genetic correlations for Holstein between milk FA and total fat estimated in Hein et al. (2018), correlations between different HFA indices and an overall fat index, only putting weight on total fat amount, were calculated. The correlation between an HFA breeding index weighting MUFA, PUFA, SCFA, C16:0, and total fat with 3, 1, 1, −1, and 10, respectively, and a breeding goal only weighting total fat is 0.91. As the correlations between the indices for MUFA and total fat and between PUFA and total fat are both near zero, then the ratio between MUFA and PUFA will be genetically stable over time using the HFA index. In contrast, using the present breeding goal for fat putting weight only on overall fat, then the correlation between the fat index and MUFA and PUFA is around −0.3, meaning that the frequency of these FA components is reduced using the present breeding goal. Furthermore, an HFA index would improve the content of SCFA and reduce the content of C16:0 because the index correlations between HFA and SCFA and C16:0 are 0.29 and −0.19, respectively, meaning that the content of total fat being SCFA is increasing and the content of C16:0 is decreasing using the HFA index. Using the HFA index would therefore maintain the MUFA and PUFA frequency of total fat at a constant level, increase the SCFA frequency of total fat and reduce the frequency of C16:0 with a cost of less than 10% lost gain in overall fat production. According to our results, this could be done without harming other traits in the breeding goal. Therefore, an easy implementation of the present results in dairy cattle breeding schemes would be to replace the traditional fat index with an HFA index in the overall total merit index without changing the relative economic weight of the index compared with other breeding goal indices.

Changing the Milk FA Profile: Effect on Technological Properties of Milk
Changing the milk FA profile can affect the technological properties of the milk. Milk with a higher unsaturated fat content has a positive impact on the technological properties of butter, such as spreadability (Bobe et al., 2007). However, increased unsaturated fat content could increase the susceptibility to lipid oxidation and thereby negatively alter the taste of butter and whole milk powder. For milk and cheese, this is not seen to be an issue due to antioxidants in the milk and the reducing environment of the cheese (Kilcawley et al., 2018). With regard to cheesemaking ability, Sanchez et al. (2018) found that fatty acids in milk were more genetically correlated with cheese yield than to the coagulation properties of the milk.

Usability of Milk Fatty Acids
Milk fat and protein can be considered proxies for assessing the energy status of the cow. A high fat-toprotein ratio is an indicator of negative energy balance (Friggens et al., 2007) and thus provides information about cows that are more prone to develop afflictions such as ketosis, displaced abomasum, ovarian cyst, lameness, mastitis, and body condition loss (Gengler et al., 2016). Milk fat, as used in the study of Friggens et al. (2007), consists of many different components. By decomposing milk fat into its underlying FA components, more information is potentially available. It has been shown that BW change in first-parity cows could be explained mainly by decreased SCFA and increased C18:0 FA in milk (Dettmann et al., 2020). Cecchinato et al. (2019) showed that LCFA and SCFA had positive and favorable genetic correlations with udder health, whereas odd-chain FA and CLA had a negative genetic correlation with udder health. This is in line with our study showing positive genetic correlations between SFA and SCFA with the EBV of the udder health index, whereas MUFA showed negative and unfavorable correlations with udder health. These results indicate that FA could be an indicator trait for disease-related traits in dairy cattle. The general health index consists of ketosis, other metabolic disorders, feet and legs disorders, early and late reproductive disorders; except for ketosis, each trait group consists of different disorders (NAV, 2018). Fatty acids might have higher correlations with some of these individual traits than to the general health index.

CONCLUSIONS
The EBV for the detailed milk FA composition in DH and DJ showed moderate correlations with EBV for fat production, but low to nonsignificant correlations with EBV of nonproduction traits. This indicates that there is room to improve the overall fat profile in milk without negatively affecting genetic gain in other important breeding goal traits, while also resulting in minor loss in overall fat production.

ACKNOWLEDGMENTS
The project SOBcows is part of the Organic RDD 2 program, which is coordinated by the International Centre for Research in Organic Food Systems (ICROFS, Tjele, Denmark). This project received grants from the Green Growth and Development Program (GUDP) under the Danish Ministry of Food, Agriculture, and Fisheries (Copenhagen, Denmark). The authors have not stated any conflicts of interest.